Instructions to use Omar401/tinyllama-alpaca-h100 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Omar401/tinyllama-alpaca-h100 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="Omar401/tinyllama-alpaca-h100")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Omar401/tinyllama-alpaca-h100")
model = AutoModelForCausalLM.from_pretrained("Omar401/tinyllama-alpaca-h100", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Omar401/tinyllama-alpaca-h100 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Omar401/tinyllama-alpaca-h100"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Omar401/tinyllama-alpaca-h100",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/Omar401/tinyllama-alpaca-h100

SGLang

How to use Omar401/tinyllama-alpaca-h100 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "Omar401/tinyllama-alpaca-h100" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Omar401/tinyllama-alpaca-h100",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "Omar401/tinyllama-alpaca-h100" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Omar401/tinyllama-alpaca-h100",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use Omar401/tinyllama-alpaca-h100 with Docker Model Runner:
```
docker model run hf.co/Omar401/tinyllama-alpaca-h100
```

See axolotl config

axolotl version: 0.8.0.dev0

base_model: TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T
datasets:
  - path: alpaca_dataset.json
    type: alpaca
output_dir: ./outputs/tinyllama-alpaca-h100

# Hardware-Specific Tuning (H100 80GB)
micro_batch_size: 32  # Max stable for H100 (tested empirically)
gradient_accumulation_steps: 1  # Disabled (VRAM fits full batches)
flash_attention: true  # 20-30% speed boost on H100

# LoRA Config (optimized for throughput)
lora_r: 8
lora_alpha: 16
lora_dropout: 0.05
lora_target_modules:
  - q_proj
  - v_proj

# Training
num_epochs: 1
learning_rate: 0.0002
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
warmup_steps: 100
max_steps: 17500  # 560k samples / batch_size 32

# Memory/Precision
bf16: true  # H100 native support
load_in_8bit: false  # Disabled (hurts H100 throughput)
gradient_checkpointing: false  # Not needed with 80GB VRAM

# Logging
eval_steps: 200
save_steps: 1000
log_format: wandb  # Optional for real-time monitoring

outputs/tinyllama-alpaca-h100

This model is a fine-tuned version of TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T on the alpaca_dataset.json dataset.

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0002
train_batch_size: 32
eval_batch_size: 32
seed: 42
optimizer: Use OptimizerNames.ADAMW_BNB with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 100
training_steps: 17500

Training results

Framework versions

Transformers 4.49.0
Pytorch 2.5.1+cu124
Datasets 3.2.0
Tokenizers 0.21.0

Downloads last month: 15

Model tree for Omar401/tinyllama-alpaca-h100

Base model

TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T

Finetuned

(109)

this model